vector computenormal(vector P, ...)
In shading contexts, computes the normal for position P using the cross product of the derivatives of P.
vector computenormal(vector P, vector N, vector Ng, ...)
In shading contexts, computes the normal for position P using the cross product of the derivatives of P. N is the original surface normal and Ng is the geometric normal. This version "adjusts" the computed normal so interpolated normals will be relatively correct.
void computenormal(int i)
(Obsolete) In SOP context, sets the hint for whether normals should be recomputed when
N change (0=never, 1=automatic, 2=always).
Functions which compute derivatives take additional arguments to allow tuning of the derivative computation.
Whether derivatives are "smooth" across patch boundaries. In most cases this is true and if extrapolation is turned on, derivative computation should be exact for C2 surfaces. However, when the VEX variables are changing with a high frequency (for example, a high frequency displacement map causing high frequency changes to the P variable), extrapolation of derivative computation may cause exaggeration of discontinuities between patch boundaries.
Adjust the magnitude of the differentials non-uniformly over patches. This will usually reduce patch discontinuities in displacement/textured shaders. However, in some odd cases you may want to turn this feature off.
N = computenormal(P, "extrapolate", 1, "smooth", 0);